Skip to content

Rebrand/shadow code - #13208

Closed
deepak-s-2000 wants to merge 5 commits into
continuedev:mainfrom
deepak-s-2000:rebrand/shadow-code
Closed

Rebrand/shadow code#13208
deepak-s-2000 wants to merge 5 commits into
continuedev:mainfrom
deepak-s-2000:rebrand/shadow-code

Conversation

@deepak-s-2000

@deepak-s-2000 deepak-s-2000 commented Aug 31, 2026

Copy link
Copy Markdown

Description

[ What changed? Feel free to be brief. ]

AI Code Review

  • Team members only: AI review runs automatically when PR is opened or marked ready for review
  • Team members can also trigger a review by commenting @continue-review

Checklist

  • [] I've read the contributing guide
  • [] The relevant docs, if any, have been updated or created
  • [] The relevant tests, if any, have been updated or created

Screen recording or screenshot

[ When applicable, please include a short screen recording or screenshot - this makes it much easier for us as contributors to review and understand your changes. See this PR as a good example. ]

Tests

[ What tests were added or updated to ensure the changes work as expected? ]

deepak-s-2000 and others added 5 commits June 25, 2026 18:30
@deepak-s-2000
feat(core): implement Ultra Token Saving with on-demand chat history
8eb45fd 
Ports shadow-chat's token optimization natively into the TypeScript core
so all LLM providers (OpenAI, Gemini, Anthropic, Ollama, etc.) get up to
97% input token reduction without any BYOK config changes.

New files:
- core/util/shadowChatSessionId.ts: stable session ID via sha256(first message)
- core/data/shadowChatDb.ts: SQLite storage with FTS5 for history/search/stats
- core/tools/implementations/shadowChatHistory.ts: 7 internal shadow_* tools
- core/llm/tokenOptimizedChat.ts: agentic loop that executes tools server-side

Shadow tools available to the LLM on demand:
  shadow_get_chat_history, shadow_search_messages, shadow_semantic_search,
  shadow_get_conversation_stats, shadow_get_tool_result,
  shadow_search_all_sessions, shadow_semantic_search_all_sessions

Modified files:
- core/index.d.ts: add ultraTokenSaving to ContinueUIConfig
- core/config/sharedConfig.ts: wire ultraTokenSaving through shared config
- core/llm/streamChat.ts: intercept calls when ultra mode enabled; block
  mid-conversation mode changes with an in-chat error message
- gui/src/pages/config/sections/UserSettingsSection.tsx: add Ultra Token
  Saving toggle in Chat settings (below Show Session Tabs)
@deepak-s-2000
Merge branch 'continuedev:main' into main
7107866
@deepak-s-2000
fix(core): correct shadow chat session scoping, promote history tools…
e7eef8e 
… to built-ins

Ultra Token Saving's session ID was hashed from messages[0], which is the
system prompt (constructMessages.ts always unshifts one) rather than the
first user message. Since the system prompt is near-identical across
unrelated conversations, this collided every chat sharing a mode into one
session, leaking chat history between them, and meant the mid-conversation
toggle guard never actually recorded a session to compare against. Session
ID now prefers the GUI's real per-conversation UUID (threaded through
llm/streamChat and tools/call), falling back to hashing the first user
message only when a client doesn't supply one.

The 7 shadow_* tools are now registered as real built-in tools
(core/tools/builtIn.ts, definitions/shadowChatHistory.ts, callTool.ts) in
the BUILT_IN_GROUP_NAME group, so they're visible in Settings > Tools and
usable by normal (non-ultra) chat, not just force-injected in ultra mode.
Ultra mode still executes them inside its own internal loop rather than
round-tripping through the client -- that's what keeps "only the current
message is sent" true -- but now yields the resolved call/result as chunks
so the UI renders the same tool-call card instead of silently swallowing
them, and streamUpdate() marks the call done+output immediately instead of
leaving it stuck on "generating" (nothing else would ever resolve it).

Also:
- Persist tool call input/output immediately at execution time for every
  built-in and MCP tool (not just shadow ones), via a hook in callTool()
  plus one in the ultra-mode internal loop. Replaces the old capture in
  saveMessages(), which only saw client-echoed results a turn late and
  never had the arguments.
- Fix shadow_semantic_search throwing on any query containing punctuation
  (FTS5 MATCH parses it as query syntax) by quoting queries as a literal
  phrase.
- Show tool arguments in the UI for tools without a bespoke display
  (previously only output was ever visible).
- core/sqlite3 and core/sharp bumped to versions with prebuilt Windows
  ARM64 binaries (dev-environment fix, unrelated to the above).
@deepak-s-2000 @claude
feat: add Claude Code CLI as a subscription-based chat provider
545ca35 
Adds a "claudecode" LLM provider that spawns `claude -p` per turn instead
of calling a metered API, so chat can run through a Claude subscription
instead of an API key. Continue's own harness stays authoritative:

- core/mcp/shadowCodeToolsServer.ts hosts Continue's own tools as an
  in-process MCP server (stateful Streamable HTTP - stateless mode is
  broken on Windows in the installed SDK) instead of letting Claude Code
  use its own Read/Write/Edit/Bash.
- core/llm/llms/ClaudeCodeCli.ts is the provider: disables Claude Code's
  built-in tools, overrides its system prompt, and yields paired
  toolCall/result chunks so already-resolved MCP tool calls render as
  history rather than pending work.
- A new claudeCodeCli/authorizeToolCall Core->GUI request
  (core/protocol/webview.ts, gui/src/hooks/ClaudeCodeCliApprovalGate.tsx)
  routes every tool call through the same policy/approval flow normal
  tool calls use.
- Cross-turn memory reuses the existing Ultra Token Saving path
  (ShadowChatDb + shadow_* tools) instead of CLI-native session
  continuity, forced on for this provider in core/llm/streamChat.ts and
  core/llm/tokenOptimizedChat.ts since it isn't optional here.
- core/tools/implementations/serverSideEdit.ts makes
  single_find_and_replace/multi_edit work when called via MCP (previously
  client-only), reusing the same deterministic preprocessArgs computation
  the GUI's client tools already depend on.

Also adds CLAUDE.md documenting the product intent and architecture for
future sessions.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@deepak-s-2000 @claude
rebrand: Continue -> Shadow Code across the monorepo
c567c80 
Rename the product for independent publishing as a VS Code extension
(+ JetBrains plugin + CLI).

Naming scheme:
- display name "Shadow Code"; package/slug "shadow-code"
- code namespace for string IDs "shadowCode." (commands, settings, context
  keys, webview view IDs)
- global data dir ~/.shadow-code; workspace dir .shadow-code/; ignore/rc
  files .shadow-codeignore / .shadow-coderc.json / .shadow-coderules
- env var prefix SHADOW_CODE_; IntelliJ plugin id com.shadowcode.plugin

Scope: extensions/vscode + gui + core + binary + extensions/cli +
extensions/intellij + docs + .github. VSIX builds clean
(extensions/vscode/build/shadow-code-*.vsix); all TS projects type-check;
core/gui unit tests green.

Icons: user-supplied spiral icon for media/icon.png; generated spiral SVG
for the activity bar (media/sidebar-icon.svg), the GUI logo
(ShadowCodeLogo.tsx), and the JetBrains plugin/tool-window icons.

Deliberately unchanged (functional / legal, not user-facing branding):
- @continuedev/* npm scope; continue-proxy provider id; X-Continue-Provider
  header; api./hub.continue.dev backend endpoints
- upstream copyright line in LICENSE.txt (Apache-2.0 requires it; a Shadow
  Code line was added)
- internal TS symbol names and webview-protocol message names
  (SerializedContinueConfig, focusContinueInput, ...)
- IntelliJ Kotlin package com.github.continuedev.continueintellijextension
- code comments linking upstream issues/PRs

Still open: marketplace publisher id (placeholder "shadow-code"); CLI binary
name (kept as "cn"). Includes pre-existing in-progress work on the Claude
Code CLI provider that was already in the working tree.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016Np5Eu8A665zkGmLuPouUn
@deepak-s-2000
deepak-s-2000 requested a review from a team as a code owner August 31, 2026 18:36
@deepak-s-2000
deepak-s-2000 requested review from sestinj and removed request for a team August 31, 2026 18:36
@github-actions

github-actions Bot commented Aug 31, 2026

Copy link
Copy Markdown
Contributor


Thank you for your submission, we really appreciate it. Like many open-source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution. You can sign the CLA by just posting a Pull Request Comment same as the below format.

I have read the CLA Document and I hereby sign the CLA

You can retrigger this bot by commenting recheck in this Pull Request. Posted by the CLA Assistant Lite bot.

@github-actions github-actions Bot locked and limited conversation to collaborators Aug 31, 2026
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Reviewers

@sestinj sestinj Awaiting requested review from sestinj sestinj is a code owner automatically assigned from continuedev/continue-code-reviewers

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@deepak-s-2000